We very much hope that you were able to join us at the Global AI summit last Thursday. We had more than 1,500 guests tune in to catch the results of the 2020 Tortoise Global AI Index, to discuss the geo-politics of AI, to examine its impact on talent, investment, and the most recent technological breakthroughs
AI is unquestionably the most transformative technology of our times, accelerating breakthroughs in health, robotics, cyber, industry, banking and more. But recognising that is only the start.
What matters is the speed and insight with which nations put AI at the heart of their policy, research and investment strategies. Economic growth, successful innovation and national confidence alike will be shaped by success in the AI space.
For all the intensity with which it is discussed in academic, industrial and governmental discourse, it is not yet seen more broadly and with true clarity for what it is: essential for national prosperity.
At the Summit, we were fortunate to have as speakers star guests such as Sir Nigel Shadbolt, chairman of the Open Data Institute, Azeem Azhar, the founder of Exponential View, Carme Atigas, the Spanish minister in charge of digitalisation and Jaan Tallinn, the co-founder of Skype, Carissa Veliz, author of Privacy is Power and Dipayan Ghosh, former tech advisor under Barack Obama among many others.
We were also on hand to talk through the latest results from our groundbreaking Global AI Index. Spoiler alert: the US took first place again, but China is fast catching up.
Here’s a readout from the day in full. You’re also very welcome to watch back any of the sessions on the Tortoise app or website. Any comments or questions please do hit reply and let us know. We’d love to hear from you. And if you are interested in becoming a member of our AI Network please register here or email Alexandra directly on email@example.com.
Opening keynote: Sir Nigel Shadbolt, founder of the Open Data Institute, in conversation with James Harding, co-founder and editor of Tortoise
- Deepmind’s success at predicting protein structures with incredible speed and precision was a huge AI breakthrough – and it was especially encouraging that Alphafold was trained on open data.
- Health and Life Sciences will be big drivers of AI development in 2021, partly (but not only) because of the pandemic. Drug development is very amenable to AI. It’s going to be “huge”, Sir Nigel said.
- Machine learning is now used everywhere, and yet national education systems have yet to adapt. Data literacy and computational thinking should be taught and embedded in the curriculum. Subjects such as history must be seen as quantitative as well as qualitative.
- The UK produces some of the best AI talent in the world. The challenge is holding on to it. The country has a great opportunity to innovate in the AI space. We have regulators with statutory powers who are thinking about innovation – including how to safely harness sensitive data for the public good.
- The centralisation of the web was something that many people didn’t anticipate. Unnatural monopolies may be beginning to emerge. Sir Nigel and Sir Tim Berners Lee are currently working on a web architecture that will “completely invert” the present centralised model. He wants data to remain in individual repositories when being accessed by third-party applications – rather than being uploaded onto the cloud. Such localisation would protect privacy and safeguard against surveillance capitalism.
Session 2: The Tortoise Global AI Index: 2020 results
- The top four countries in the 2019 Index retained their position in 2020’s Global AI Index. But China (2) narrowed the gap further on the US (1), while Canada (4) almost knocked the UK off its bronze-medal position.
- France and Germany – both in the top 10 – are starting to nip at the heels of the UK, which is slipping back thanks in part to a government strategy not as robust as Canada or the Netherlands.
- The biggest risers on the Index – which ranks more than 60 countries according to their AI development capacity, were Israel (5), the Netherlands (7) and Finland (11). Israel had a particularly good 2020: investment hasn’t stopped and talent remains excited despite the pandemic. It has the highest number of AI startups per capita.
- Russia (31) came quite far down in the Index but it was disadvantaged by its lack of transparency: much of its AI spending may be going to defence – where we can’t track it.
- China is trying to drive up its talent pool – including by attracting PhD and Postdoc researchers back from the US. The country is also trying to develop its chip manufacturing base, a traditional disadvantage. But Beijing still lags behind the US in terms of private investment.
Session 3: The geo-politics of artificial intelligence: revisiting the New World Order, with Azeem Azhar, founder of Exponential View, Carme Artigas, Spanish secretary of state for digiliatlision and AI, Benedict Evans, consultant and long-time analyst, and Casper Klynge, former Danish Tech ambassador now VP of European government affairs at Microsoft
- AI has the power to transform industries, economies and in the military. It’s a general purpose technology that could rewrite rules of society in the same way as electricity or the steam engine.
- Technology clusters are critical for getting ahead. Silicon Valley has been joined by Montreal in Canada, the “golden triangle” in the UK, and in the big conurbations in China. Clusters are difficult to copy.
- To succeed in the AI race countries need to overcome the following challenges, said Carme Artigas: access to data, access to computer power, access to financing and, above all, access to talent.
- AI is a “deeply unhelpful term”, according to Ben Evans. Like relational databases that allowed companies to store and manage data in new ways and transformed companies such as Zara and Apple, AI is just one of many enabling layers developed by technologists, he said.
- Most countries have a sense that they are in a geopolitical AI arms race and that some regions are trailing behind, Casper Klynge said. Europe feels squeezed between a rock and a hard place. Casper said it was also important to examine what the AI’s arms race means for South East Asia, for Latin America and for Africa.
- For Carme, countries need AI strategies because the technology has so many wider implications on society. Europe needs to develop AI according to its principles and values, she said. “In the US data belongs to companies, In China it becomes to governments, in Europe it will belong to citizens”.
Session 4: AI’s ugly underbelly: winners and losers of the workforce, with Christine Foster, CCO, The Alan Turing Institute, Jean-François Gagné, CEO of ElementAI, Taavi Kotka, first CIO for the Government of Estonia, and Frida Polli, CEO of Pymetrics
- The AI talent pool is growing, but not consistently. The number of pure AI researchers is actually decelerating while the real expansion is in jobs that apply AI techniques. The ecosystem is becoming more mature.
- China and the US are still the biggest users and attractors of AI talent – by far, said Jean-Francois Gagne. Everyone else is clumped in the middle. The level of maturity of the AI ecosystem is a key driver to attracting talent.
- To get the best talent recruiters have to look beyond traditional pipelines and look at soft skills that are more equally distributed among the population than hard skills, Frida Polli said. Algorithmic recruitment techniques can actually help with this.
- Some AI workers – on Amazon MTurk – are paid very low wages in practice. However, said Christine Foster, there is an increasing acknowledgment that it takes a team to build a good AI product – and that in future AI workers will be on a more equal footing. Christine pointed out that digital platforms make exploitation much easier to see than previously – making a regulator’s job easier. The Turing Institute is working on eradicating modern slavery through modern data sets, she said.
- Estonia has attracted AI talent in spite of its size because workers can see and feel its digital society, Taavi Kotka said. The public and private sectors exchange data and see things in the same way: it’s interesting to try out new skills. Plus Estonia’s education system is the best in Europe (although its weather is not).
- The effects of the pandemic have not been even. It’s made some people much less productive. It’s also had a disproportionate effect on women: they have been unfairly impacted and this is going to have a significant impact going forward.
- Jean-François pointed out that an “ever growing” number of M-Turk style jobs will eventually be eliminated by AI advancements – potentially creating another type of problem.
Session 5: AI you can trust: taking ethics from rhetoric to reality? With Lofred Madzou, Project Lead for AI at the World Economic Forum, Chris Wigley, CEO of Genomics England, Dipayan Ghosh, former tech advisor under Barack Obama, and Carissa Veliz, associate Professor in Philosophy at the Institute for Ethics in AI
- Chris Wigley talked about the data protections set up to safeguard the Genomics England database. Researchers could examine the data but not bring it onto their own laptops. Another safeguard was a veto-power that a panel could use to prevent a researcher accessing the data. Chris pointed out that sick people and well people have very different attitudes to their health data. He had experienced “aggression” from healthy people who object to his use of data, but not from sick people whose data it was.
- To ensure ethical behaviour around AI, certain behaviours should be banned by law, according to Carissa Veliz. These included buying and selling data. Personalised content is polarising society, Carissa said.
- Third party audits would go a long way to solve algorithmic biases, Lofred Madzou said. Without their intervention we won’t know the scale of any algorithmic problem as the data is in the hands of private companies. So we need academics, activists, and businesses to work together to audit these algorithms. Collaboration is as important as the audits themselves.
- There’s a tendency to conflate “true” AI and a basic algorithm, Andrew Girdwood, a Tortoise member, said. Regulating a neural network is like regulating someone’s brain, he said – almost impossible. Dipayan Ghosh said Google and other companies often made this argument to regulators: that self-learning algorithms were too complex to examine. One way of regulating such algorithms effectively would be to look at inputs and outputs rather than the underlying code, he said.
- Algorithms can work well in the lab but can be disasters in real life, Ms Veliz said. AI algorithms can have as big an effect as drugs yet they undergo no robust testing. They should undergo randomised controlled trials before being released, she said.
Session 6: AI: the breakthroughs that will shape your life, with Jaan Tallinn, co-founder of Skype, Sophie Hackford, Futurist and researcher; Siraj Khaliq, Partner on the Investment team at Atomico, Yasir Khokhar, founder and CEO of Connecterra; and Timothy Revell, Comment and culture editor at New Scientist
- Jaan Tallinn always got excited about Deepmind’s breakthroughs, he said. But at the same time, he thought: “wait” – there might be a downside to such developments. One reason he’s focusing on the risks of AI is that the upsides themselves are so big.
- One practical AI breakthrough is in farming, Yasir Khokhar told us. His company relies on AI-driven sensors to increase yields and farming efficiency by 20% and reduce antibiotic usage by about 50%. AI is helping identify patterns that humans can’t see. Yasir’s sensors – placed on livestock – can predict whether a dairy cow is going to be sick a few days before it gets ill.
- On the consumer side, the number of AI sensors in houses and cars will only increase, said Sophie Hackford. So-called smart glue will be woven through our homes. We might even buy a Google house in the not too distant future.
- Translating breakthroughs like Deepmind’s protein folding AI into practical solutions might be very hard in practice, however. GPU performance is increasingly far more slowly than AI models, Siraj Khaliq said. Five years ago a state of the art model for AI was 25 million parameters. This year GPT3 has 175 billion parameters. During the same period, computers have “only” gotten 10 times faster. We need a computational breakthrough, Siraj said.
- Other sectors where we’re seeing AI breakthroughs include health and biotech, therapeutics and fintechs – but almost every business is going to use AI. Augmented intelligence is also a crucial potential breakthrough: by analysing our behaviour, AI could coach us in how to be a better employee – even a better parent or spouse.
- Alison Lowndes, an AI professional and an attendee, gave an industry perspective. She said she wasn’t worried about any developments and that any mistakes were rapidly remedied.
- Timothy Revell, a New Scientist journalist who recently interviewed Deepmind’s Demis Hassabis, said AI companies were increasingly relying on “synthetic data” – and did not have to rely on someone’s actual data to develop AI systems.
Some further reading…
Gemma Milne’s new book Smoke & Mirrors shows us how to separate the truth from the hype around nine cutting-edge areas of science and technology.
The Global AI Talent report by Jean-François Gagne, Simon Hudson and Yoan Mantha.Tabitha Goldstaub’s How to Talk to Robots: A Girls’ Guide To a Future Dominated by AI explores both the risks and benefits to AI, whilst spotting gender bias along the way.